Artificial Intelligence and its role in the Armed Forces

?Artificial Intelligence and its role in the Armed Forces

?

Section One:? Foundations, Tools, and Methodologies of Artificial Intelligence

?

Introduction

?

Artificial intelligence (AI) refers to the technology that simulates human mental abilities such as the ability to learn, create, and react. This technology is now producing tools and new capabilities that would potentially be beyond human control. This type of technology is being integrated in practically all areas of human endeavor. [i] Militaries are developing AI systems that support combined operations, joint engagements, and other command activities. These include the use of lethal force. Some applications can produce safety risks or dynamics that can impact the arms race. Generative AI, an advanced category of the technology where AI begins to take on increasingly human-like characteristics, could have tremendous impact on military training and educational programs. It can facilitate very advanced and comprehensive training environments, including some as complex as military surgical simulations. AI can enhance effectiveness across a broad array of military operations, programs, and activities from the use of lethal weapon systems and drone swarms in high-intensity warfare to leveraging algorithms to fuse data and realize virtual connections in the information domain. AI can manage vast quantities of data and, if properly programed, use trend analysis to identify patterns humans might miss.? Artificial intelligence is now used by information systems to play games such as chess that require advanced intelligence. It can leverage visual perception, speech recognition, and language translation to support human endeavors in increasingly complex ways. With its increasing prominence, AI is demonstrating its usefulness more intelligently every day in all areas of life, leading innovations in areas as diverse as security services and public speaking.

?

Recently, there have been a variety of categories of AI receiving a lot of media coverage. One of the most significant is advances in natural language processing (NLP), where computers can understand, interpret, and generate human language. This type of AI can automate language-related tasks, accelerate focused data analysis, improve communications, and enhance individual, team and organizational effectiveness. Another area of significant advance is in Artificial Super Intelligence where AI surpasses human intelligence and by using system instructions write new computer code facilitating independent systems advancement. Such technology could research and develop new areas of capability with accuracy and precision and use these to understand and meet requests for specialized communication such as texts or images. An area that has been incredibly popular recently is computer vision, which has new image analysis features and can produce amazing results through its ability to understand and interpret visual information. In addition, there have been experiments in using artificial intelligence that innovates. These innovative systems might offer novel and creative solutions to existing challenges and identify unforeseen opportunities. ?While impressive advances have been made that have demonstrated the potential of AI there is still opportunity for additional new applications for AI. It is important that militaries consider how AI can help them to be more efficient and effective.

?

The United States Military is one organization that has used artificial intelligence in many areas, even before AI became widespread in civilian life. With the input of correct data, AI can improve performance in combat and non-combat operations, programs, and activities. AI is currently being used in many different functions in the military. Over time, various industrial technologies will leverage AI to more safely perform even the most dangerous jobs, almost eliminating the need for human input in some situations. But ethical issues and technical limitations mean that a human factor is still required when implementing AI systems.

?

?Foundations of Artificial Intelligence

?

Artificial intelligence (AI) generally refers to technology that attempts to simulate human intelligence. It uses data to replicate human cognition such as skills like learning, problem solving, pattern recognition, and language. It includes subfields such as machine learning, expert systems, robotics, and computer vision. It is proving useful in a wide variety of industries including defense.

?

Coursera, an organization that produces online learning content, provides an excellent introductory course on AI.[ii] ?It discusses new AI tools gaining rapid usage including ChatGPT, Google Translate, Tesla computer vision. For example, it shows how Netflix uses machine learning algorithms to create personalized recommendations based on a viewer’s past viewing history. The course teaches what AI is, what it does, what types of AI exist, and what risks AI might present to society. AI is already integral to the modern world. It is being used in the workforce to reduce costs and physical and financial risks. Surgeons are using it in delicate operations and to reduce infections. According to Coursera, there will be a point where AI achieves or exceeds human intelligence, and we will soon be unable to reliably identify if we are interfacing with AI of a human. AI might soon achieve sentiency which basically means consciousness or self-awareness. Possibly, AI will be capable of advanced emotions such as empathy or anger. This raises challenging ethical issues. So, while AI provides tools with great promise, it also represents great risk. AI tools present both benefits and dangers. They offer greater accuracy, reduced costs, personalized services, better decisions, and enhanced creativity. But they also threaten jobs, discrimination, cybersecurity risks, reduced transparency, and misinformation. The use of these tools must consider these and other ethical issues.

?

?

?

?

Tools of Artificial Intelligence

?

According to Professor Arend Hintze of the University of Michigan[iii], there are four types of AI: reactive machines, limited memory, theory of mind, and self-awareness. Reactive learning machines are tools that perform a narrow scope of tasks such as playing chess. They possess no prior knowledge and only react within a limited set of logic and commands. ?It is the most basic form of AI. Limited memory machines represent the next level of AI capability. Limited memory machines can store data about past events and use this memory to interact with the world. Self-driving cars, a potentially revolutionary tool that could transform modern human society, are an example of this type of AI. Theory of mind machines, still mostly theoretical tools that will be able to create representations of the world, are just now emerging. Chatbots, and virtual assistants such as Siri could be seen as early theory of mind machines. Self-awareness machines, seen as the ultimate form of human developed AI, are potential tools theoretically possessing something like human sentiency. These tools do not currently exist and some experts question if this level of AI is even possible.

?

From its humble beginnings in the 1950s with the development of “expert systems” that emulated human experts in specific fields, through narrow AI tools with specific skills and limited context, to general AI that is possessing increasingly advanced cognitive skills, AI is proving to provide tools of great value to humanity. But issues such as bias, misinformation, and liability will all need not be considered as AI becomes more predominate in our lives. The type of information an AI tool provides is constrained by the data and programing logic human designers and developers provide. As AI becomes more advanced and provides responses to increasingly complex situations, these constraints will produce more biased information. This will be particularly challenging as human programmers begin to modify AI response to what they perceive as moral or correct. This will result in increasing levels of misinformation. As AI becomes more advanced and provides output on vastly greater amounts of imputed data, humans might be unable to confirm the veracity of AI response. One reason self-driving cars have not advanced more than they have is liability. Who pays when a self-driving car causes injury or property damage? The shortcomings of AI will inevitably emerge and in the life-and-death arena of military operations, they will need to be thoroughly considered so leaders can make informed decisions regarding the use of AI tools within acceptable levels of risk.

?

Methodologies of Artificial Intelligence Usage

?

The methodologies of AI usage encompass a wide variety of activities. These methodologies help the development, implementation, and assessment of AI tools. They are key to designing the best AI solutions, developing models, evaluating performance, and guaranteeing effectiveness. Among these methodologies are machine learning, deep learning, nature language processing, computer vision, data processing, model validation, interpretability, and continuous learning adaptation. These methodologies allow optimal AI solutions to address complex tasks with better decision making, automation, and innovation enhancement.

?

The military is looking for ways to leverage AI. Sentient Digital, Inc. provides an extensive review of the most useful applications of AI by the military in 2024.[iv] These include:

?

1. Warfare systems: AI can make various systems more efficient and less reliant on human input. This means less maintenance and fewer human errors. Humans can focus on essential tasks. Specifically, the Pentagon recently updated its autonomous weapons policy to make it safer and more ethical while creating the required workforce.

?

2. Drone Swarms: An exciting emerging military application of AI involves drone swarms leveraging AI. ?AI-controlled swarms of drones are programmed to act in the same manner that swarms of insects act in nature. They can communicate the distance, direction, and elevation of a target, as well as any potential dangers. The side better able to leverage AI-powered drone swarms towards military objectives will win in future conflict.

3. Strategic Decision Making: AI’s algorithms can collect and process data from numerous different sources to aid in decision making, especially in high-stress situations. Ideally, AI and humans will work together, combining human ethical understanding and AI’s analytical abilities. AI can also create simulations to test out possible scenarios, allowing for more informed decision making. With close human supervision, generative AI can enhance strategic thinking.

?

4. Data processing and Research: Processing large amounts of data is time consuming for humans. AI can be helpful for quickly filtering through data and group information from various datasets. Military personnel can create plans based on a more complete picture of the situation.?Generative AI’s analysis capabilities can find connections that might escape humans’ notice. AI models can communicate in a variety of languages. It can analyze vast amounts of news and social media to identify new information. Generative AI can also speed up the analysis process and generate and compare thousands of scenarios by making small changes to variables.

?

5. Combat Simulation: These have been used by the military for many years. By combining software engineering and computer science to build digitized models that prepare soldiers with combat systems deployed during operations the military can provide essentially a virtual “wargame” that is used to train soldiers.?The use of simulations in training is extensive and will be covered in Section 2 of the chapter.

?

6. Target Recognition: AI can improve target recognition and allow defense forces to acquire a detailed understanding of an operation area by examining information more quickly than humans can.?With generative AI, there can be a two-way discussion about this information. AI systems can assess conditions in the operational environment and save time and human resources, speeding up the targeting process. But as always, humans should make the ultimate targeting decision.

?

7. Threat Monitoring: Situation awareness requires operations that gain and analyze information to aid in military activities. There are unmanned systems that use AI to aid monitoring threats. They can monitor border areas, recognize threats, and alert response teams. Additionally, they can strengthen the security of military bases, as well as increase the safety of soldiers in combat.

?

8. Cybersecurity: Due to its nature as a form if information system, AI provides superior advantages in the arena of cyber security. AI can protect programs, data, networks, and computers from unauthorized persons. AI also has the skills to study patterns of cyber-attacks and form protective strategies to fight against them. Generative AI’s analysis, scenario generation, and communication capabilities improves cybersecurity in military settings. But functions such as the ability to write malware may make AI dangerous in the hands of bad actors. It is critical for the military to have access to the most advanced and tailored AI cybersecurity solutions.

?

9. Transportation: AI can be used to optimize logistics. This lowers transportation costs and reduces the need for human input by, for example, plotting the most efficient route to travel under current conditions. Self-driving vehicles will dramatically reduce the risk of distributed logistics in combat zones.

?

10. Casualty Care and Evacuation: AI in a highly emotional environment can help with analyzing the situation and suggesting the best course of action for casualty evacuation. But AI does not understand the emotional and contextual factors involved in a life-or-death situation. AI uses algorithms and large medical databases to provide indications, warnings, and suggestions for treatment. It can provide rapid analysis to give humans more information on which to base their decisions.

?

As the list above shows, AI is a multifaceted arena of technology that affects a wide array of human endeavors. But all share a commonality: they all provide tools that make man more efficient and effective in accomplishing tasks. The name artificial intelligence communicates attempts to emulate human intelligence. To the degree that AI proves to be a beneficial tool to humans, its development will continue. The benefit will be defined through a term now common in the literature concerning AI: Human-Machine Teaming. The article “How Modern Militaries are Leveraging AI”[v] investigates the implications of incorporating AI into combat operations. It does this by focusing on human-machine teaming (HMT). It concludes that HMT has the potential to change the character of war by “increasing situational awareness, improving decision-making, extending the range and lethality of human operators, and gaining and maintaining advantage across the multi-domain fight” while increasing administrative and logistical efficiency and reducing costs. The article emphasizes the need to see HMT as more than humans working with uncrewed systems such as drones to include non-physical systems such as decision support software. Much of the promise of the currently conceptual Joint All-Domain Command and Control system[vi] being developed by the United States Military to “seize, maintain, and protect information and decision advantage over adversaries” will only be possible through some of the applications described above, such as Strategic Decision Making, Data Processing and Research, Target Recognition, Threat Monitoring, and Cybersecurity. Culture, bureaucratic, and human confidence challenges will need to be overcome to see the full benefits of HMT come to fruition. The experimentation, training, and systems development required to maximize HMT success are discussed in Section 2 below.?

?

The article “How Modern Militaries are Leveraging AI” also provides a review of several salient HMT initiatives. One frequently envisioned is the idea of the “loyal wingman” in which uncrewed systems operate alongside crewed systems extending human operational reach and expanding capabilities. Such uncrewed systems can assist in information gathering, targeting, and engaging enemy assets while often reducing human operational risk and expanding operational endurance. ?Nonphysical information systems can now use computer vision technology to scan huge amounts of both text and visual data looking for risks and targeting opportunities, both informing human actors and facilitating better targeting decisions. HMT may provide the critical edge needed to penetrate increasingly complex and effective anti-access and area denial environments modern militaries such as those of China, Russia, and Iran are fielding. In 2021, the United States Secretary of the Air Force confirmed[vii] that the Service intended to use AI to “significantly reduce the manpower-intensive tasks of manually identifying targets—shortening the kill chain and accelerating the speed of decision-making.” He also confirmed that AI had already been used in an operation kill chain and that AI is “shortening the kill chain and accelerating the speed of decision-making.” At present, AI is supplementing and not replacing the human element in the kill chain. The final decision to kill or not is still in the hands of humans. But it is likely that this kill chain will soon become autonomous and the ethical issues involving machine-driven lethal target engagement will need to be addressed.

?

AI will also provide great benefits to the intelligence services. AI-enabled software can be used to support human-authored intelligence assessment products. It can review human reporting for accuracy and make suggestions on ways to expand or improve these products. AI is experiencing rapid development. The most significant development in AI has been the wide public availability of Generative AI. Improvements in large language models (LLM) permit applications like ChatGPT to receive prompts and create responses. Since humans can easily understand these responses, these models promise to speed up the intelligence cycle. These include applications such as GPT-3[viii] and PaLM[ix] that represent significant milestones in the improvement of LLM tools. These applications are surprisingly human-like and can now learn from human interaction. These are applications that can now be categorized as early “theory of mind” systems. A single AI system can now receive and interact with inputs formatted as text, images, audio, and video together. Photorealistic images are now generated using text alone. All this is allowing for the production of a broader array of intelligence products. Humans can increasingly interact with AI in a similar way to the way they interact with other people, facilitating more rapid and precise reporting. This HMT will dramatically expand the quality of intelligence but there are also serious security issues presented by this emerging technology. Accuracy, verification, privacy, and target legitimacy concerns may arise. The military continues to study these technologies and integrate them into operations as efficiently and effectively as possible.

?

Section Two: The role of artificial intelligence in developing and training qualified Armed Forces

?

Introduction

?

As Section One explains, AI is developing at a rapid pace. Organizations and enterprises across society are using AI tools to increase efficiency and effectiveness. The military can benefit from the opportunities of AI. But it is also vulnerable to the risks AI presents in unique ways. Optimally and safely integrating AI tools into military operations, programs and activities will require education and training. Many of these endeavors will be iterative in nature with the military learning through training and education and developing new capabilities in the process. Done correctly, the military will become aware of the challenges and opportunities AI creates. These will drive the development of new capabilities. Experimentation through exercises, war games, simulations and other training and education will inform this process. All militaries most appreciate that their adversaries and potential enemies are also working to leverage the power of AI. Future battlefield victories, and defeats, will hinge on the outcome of these competitive efforts. The military that dominates AI will dominate the future battlefield.

?

The military uses AI in training and development for a variety of purposes to enhance efficiency, effectiveness, and realism. Overall, AI in military training helps optimize resources, improve learning outcomes, and better prepare personnel for complex and dynamic operational environments. Done correctly, AI can even make training exciting and entertaining. A common term in AI training and development is gamification. By transforming learning into a competitive game, trainees are not only motivated but eager to dive into new knowledge and skills. Whether it’s through simulations, quizzes, or interactive videos, gamification adds an element of entertainment and challenge to the learning process. AI can unlock a sense of individual and team achievement as trainees reach new levels of proficiency in what they see as a competitive challenge. It taps into individuals’ competitive spirit and desire for accomplishment. With the popularity of gaming, it’s no surprise that turning training into a game is becoming a go-to strategy in training and development. Well-designed AI training can motivate employees through game-like learning experiences. It increases engagement and motivation to learn. And through a variety of different formats like quizzes, simulations, and interactive videos, training can stay interesting and adaptive.

?

Current ways AI is being Utilized in Training

?

AI is being integrated into training in a growing variety of ways. The following provides a list of common types of AI currently being utilized. It is important to appreciate that all of these ways of utilizing AI in training work best in a human-machine team (HMT) context.

?

1. Virtual Reality (VR) and Augmented Reality (AR) Training: AI is used to create immersive training environments through VR and AR to simulate real-world scenarios and provide realistic training experiences while providing hands-on training in a risk-free environment. Consider a scenario where soldiers are undergoing training to prepare for the hazards of the combat environment. Generative AI, combined with VR, can simulate realistic scenarios, allowing learners to navigate through potential dangers in a controlled, virtual environment. This not only enhances skills retention but also provides a risk-free space for hands-on learning. AR applications powered by generative AI can overlay relevant information in the real world, creating a seamless blend of virtual and physical environments. This is particularly valuable in technical training, where learners can receive step-by-step guidance or access additional information by simply looking through AR-equipped devices.

?

2. Simulation and Decision Support: AI algorithms are used to create simulated environments where trainees can practice decision-making skills and improve critical thinking and strategic planning. AI will enable vastly more rapid and better decision making in combat (e.g., target selection) at all levels of war and combat support (e.g., resupply priorities). In their 2017 article?On Hyperwar,[x] retired USMC General John Allen and SparkCognition corporation founder Amir Husain discussed the concept of hyperwar which diffuses?the traditional “observe, orient, decide, act” decision loop developed Colonel John Boyd, USAF, and expands monolithic decision-making structures into resilient, scalable, adaptable?“kill webs”, with the automated discretion to select, target, and engage with opposing forces faster than their human counterparts.?But simulation studies have since shown that AI cannot replace humans completely in decision making. Instead, it will complement humans making limited decisions with pre-established algorithms that will need to be defined, developed, and rehearsed in training. In time, as these algorithms will fuse all domains (land, sea, air, space, information) in ways humans will never be able to do and facilitate decisions at a vastly more rapid rate. In military training simulations, AI plays a crucial role in creating realistic and immersive environments for trainees to practice various scenarios. Here are some ways in which AI is used in military training simulations:

?

Behavioral Modeling: AI algorithms simulate the behavior of enemy forces, civilians, and other entities in the training scenarios. These models adapt and respond to trainee actions, providing dynamic and realistic challenges. These models can help influence behaviors through the development of skills and knowledge competencies in areas such as decision making and forecasting. Through behavioral modeling, training and development programs can integrate agent and systems dynamics. Markov models can be considered a subset of behavioral modeling. Markov models relate independent and dependent variables and help predict the sequence of training activities that produce the best learning outcomes. Behavioral models can also inject game theory into training programs that can be used to predict the outcome of trainee decisions. Neural networks are emerging behavioral models that attempt to emulate human reactions and can provide feedback to trainees and program supervisors although these networks are still in need of development and have limited applicability without risking faulty feedback. In summary, behavioral models can enhance learning outcomes by injecting human-like feedback into training programs.

?

Scenario Generation: AI generates diverse scenarios based on a wide range of variables, such as terrain, weather conditions, and enemy capabilities. This helps create a variety of training experiences to sharpen different skills. It is particularly beneficial in training designed to improve decision-making as the scenario in which a decision must be made can be rapidly modified.? Of course, as with all AI, the quality and reality of a generated scenario is dependent on the data inputted into the system and programming algorithms.? As more complex and complicated scenarios are created, the level of scenario accuracy and completeness will suffer.

?

Decision Support: AI systems provide decision support to trainees by analyzing their choices and outcomes, offering insights into alternative strategies and potential consequences. This helps improve decision-making skills under pressure. As with all AI, the quality of decision support in training depends on the accuracy, relevancy, and completeness of data injected into the system. As more complex, cognitive decisions are required that cannot be quantified and become more subjective, this type of AI becomes less valuable and requires greater human involvement. ?In the military, decision making tends to be a group activity. Whether the challenge is scientific and quantifiable, like employing air power in joint operations, or situational and qualitative like operating in continuingly evolving operational environments where the human factor is salient, AI can help human teams and leaders make better decisions. Using AI to promote rapid and adaptive decision making is a key strategic goal of the US Department of Defense Data, Analytics, and Artificial Intelligence Adoption Strategy published in 2023[xi]? which identified accelerating decision advantage as its overarching purpose.

?

After-Action Review: AI captures and analyzes data from training simulations to generate after-action reports. These reports highlight performance metrics, lessons learned, and areas for improvement, enabling trainees to reflect on their performance. Although such feedback can assist reviews, it will require human confirmation and extensive development efforts before being integrated into a military training simulation. This is an expensive and time-consuming effort. As training moves from the technical / tactical to the operational and strategic levels of military operations, the value of AI in after-action reviews will become more subjective and inaccurate. Human judgment will be increasingly required. But this form of AI, if properly employed and adequately developed, promises to speed up and improve the after-action review process. The greater the complexity and completeness of data input, the more valuable and comprehensive such tools will prove to be in assessing training results at all levels of warfare and in all operational environments.

?

Dynamic Adaptation: AI can modify a training methodology or program sequence to meet specific trainee situations. If a trainee is slow or weak in an identified area, dynamic adaptation can adjust the experience to reinforce learning outcomes. Adaptations can be based on trainee actions and performance, providing tailored challenges and learning experiences. This ensures that training remains engaging and effective for each individual. Through dynamic adaption, both the quality of learning outcome and the actual learning experience can be continually improved as the feedback can inform both the program execution and future program improvements.

?

In summary, by leveraging AI in military training simulations, the armed forces can enhance the realism, complexity, and effectiveness of their training programs, preparing personnel for a wide range of real-world situations and promoting continuous improvement in skills and decision-making.?

?

3. Performance Analysis: AI is used to analyze and evaluate the performance of trainees during training exercises. It provides feedback on individual and group performance, allowing for targeted improvements. One company, Cerego, is already actively engaged with United States Military customers in this arena. Instructors can load training content into the company’s software system[xii] and its AI automatically produces tailored learning experiences, schedules, and assessments. Cerego claims that their learning system can help instructors identify learners who are struggling early on, before they take a certification test or go on deployment. Program leaders can use Cerego’s performance analytics tools to receive real-time, actionable dashboards and reports that track individual and team progress. Instructors and learners receive real-time, visual feedback on progress, content difficulty, study time, etc. Program leaders gain an in-depth understanding of learners’ cognitive and behavioral attributes, including agility, diligence, and knowledge metrics.

?

4. Personalized Training Programs: AI can customize training programs based on the individual strengths and weaknesses of each trainee. This personalized approach helps optimize training outcomes. It also ensures all trainees have a common knowledge base allowing them to participate more effectively in training. Finally, Cerego claims that using its system can save time and money by increasing the speed at which teams learn and retain information. Employing mobile apps such as smart phones, that today’s youth are very comfortable with, learners can study anywhere and at any time and retain knowledge and skill in significantly less total study time than using traditional methods. During one use of Cerego, soldiers who used the software for an average of just one minute per day saw a 13 percent increase in their test scores. Those who averaged 3.5 minutes per day saw a 26 percent increase in their scores.

?

5. Predictive Analytics: AI algorithms are used to analyze training data and predict future performance trends. This allows trainers to identify areas that need improvement proactively. Used in training, predictive analytics can forecast how effective a soldier will be in a real combat scenario based on their performance in a virtual environment. AI can also be integrated into training to determine how combat systems will function in combat. For example, the AI company Uptake received a $1 million contract in 2018[xiii] facilitated by the Defense Innovation Unit to integrate the company’s predictive maintenance software into Bradley Fighting Vehicles. This promises to increase systems readiness rates.

?

6. Autonomous Systems Training: AI is used to train military personnel in operating and interacting with autonomous systems such as drones, unmanned vehicles, and robotic platforms. The war in Ukraine has highlighted the need to develop such training programs. In June 2024, the Ukrainian military launched its Unmanned Systems Forces (USF)[xiv], a new branch of its military focused on drone warfare that officials said is the first of its kind. This force will help Ukraine prepare for the war of the future. The United States will be one of the main strategic partners of the USF and will help facilitate everything from reconnaissance to drones production. The USF will also collaborate with Latvia, Great Britain, Sweden, Estonia, Germany, Netherlands, Lithuania, Denmark, Canada, and Australia in the development autonomous systems tactics, techniques, and procedures.

?

7. Adaptive Learning: AI systems adapt training scenarios based on the performance of trainees, ensuring that they are challenged at an appropriate level to facilitate skill development. The Cerego learning system discussed above uses adaptive learning algorithms that are based on proven cognitive science principles and AI-driven machine learning to automatically generate a personalized study schedule that eliminates the cognitive burden of knowing what and when to study.

?

8. Medical Training Simulations: AI-powered medical simulation technology is used to train military medical personnel in realistic scenarios, improving their preparedness for combat situations. In one case, the United States Army used the Cerego learning system during the Tactical Combat Casualty Care to be cut the length of the course from 12 to six hours.

?

9. Cybersecurity Training: AI is utilized to simulate cyber threats and attacks, preparing military personnel to respond to a range of cybersecurity challenges effectively. AI exists in the information domain which increasingly is linked to computer systems, networks, devices, and data that are vulnerable to cyberthreats. AI can improve cybersecurity training by augmenting existing tools. By its nature, AI can be used to identify cyber threats. It can create parts of training programs and allow trainees to respond to cyber threats in a controlled environment. AI can use training analytics to improve cybersecurity training. It can also be used to automatically share best practices among relevant organizations involved in cybersecurity.

?

10. Language and Cultural Training: AI tools are employed to provide language training and cultural awareness programs to military personnel deployed in diverse regions around the world. As “theory of mind” AI systems advance and AI can better understand and interpret human emotions, intentions, beliefs, and motivations, AI will become more valuable in this arena of training. The ability of information applications to interpret human communication in the form of words or speech and translate this communication from one language to another without corrupting meaning improves, AI will become more valuable in supporting multinational, cross-cultural operations, programs and activities. Recent developments in neural machine translation algorithms have made AI language and cultural training applications more accurate and effective. These applications have developed to the point where comparative analysis can be conducted. For example, in one recent study comparing the quality of translations from English to Arabic[xv], 59 undergraduate translation students were required to translate a text from English into Arabic.?One group translated without the support of AI while another used Google Translate and a third used the translation application Reverso Context. Translations were evaluated and scored using a rubric developed by the American Translators Association’s Certification Program. The Reverso Context application performed significantly better, demonstrating that there is now a quality difference in this AI that can be leveraged to improve such training.

?

Artificial Intelligence in Wargaming, Simulations, and Experimentation in Training

?

????????????? AI can benefit force development in ways beyond individual training. It can also benefit force development efforts. These often overlap with individual and collective training. Preparing military forces for operations in all domains and environments across the spectrum of conflict is a multifaceted, complex endeavor.

?

In “Artificial intelligence and its role in Training and Development” [xvi] the role AI is playing in revolutionizing employee training and development is discussed. As mentioned above, AI allows customized feedback that can be adapted to each individual’s unique requirements. Performance can be monitored for trends and behaviors that identify individual knowledge and skills gaps. AI assisted assessments can identify individual strengths and weaknesses and help produce unbiased recommendations for unique development plans. Training plans can be modified to individual’s preferred learning methodologies. Feedback can be immediate and continual.

?

The article categorizes the role of AI in training and development into five subjects. Learning algorithms allow programs that can be adapted to individual needs. Performance analytics provide immediate and continual progress and improvement feedback. Cognitive coaching provides personalized feedback and focuses on unique challenges. Adaptive assessments allow feedback to be adjusted to current skill level as the individual improves performance. Finally, visual simulations allow for a safer and less expensive learning environment. AI is improving training outcomes in many industries. All five of these are reflected in current military training and development, which is a vast and complicated enterprise. Because of the complexity of military operations, training developers must consider the huge number of training methodologies and tools available and shape these into the most efficient and effective training experiences possible. Three methodologies often used by military are wargaming, simulations, and experimentation. Each of these methodologies is considered below. ?

?

Wargaming: Wargames are a time-honored way for military professionals to see how concepts of operations, tactics, or new technologies might play out in the real world. Wargames attempt to simulate reality in a way that allows human observers the ability to observe, develop, and assess action.? This action is usually quantitative and can thus be measured. Thus, AI will provide extremely beneficial to good wargaming. But AI is only as good as the data it receives. Wargames are similar to simulations but differ in that wargames seek to simulate and understand human decision-making dynamics of adversaries, allies and other key stakeholders in specific scenarios while simulations seek to imitate the functioning of a process or system in the real world. Thus, while similar, the salient difference is in the cognitive domain. This paper will discuss the use of wargaming in training and development first and subsequently discuss the use of simulations.

?

There is growing interest in injecting AI in military wargames. However, the results have at times been frustrating. For example, RAND published a study on the use of AI in wargaming[xvii] in January of 2024 that concluded that AI is far from achieving human level intelligence. It cannot replace humans in wargames, nor can it apply human judgments. It recommended that AI only be used in certain areas of wargames conducted for certain purposes. According to the study, AI is useful at the tactical level, where algorithms can be written that replicate the dynamics of weapons systems and effects of the operational environment. But AI cannot currently replicate human imagination and thus is far less effective strategic level wargames.

?

These concerns are echoed in a paper published by the Centre for Emerging Technology and Security[xviii] located in the United Kingdom. This research looked at ways AI could increase wargaming efficiencies and improve human decision making. It identified 15 uses of AI in wargaming and considered the envisioned benefits and challenges of each in wargame design, execution, analysis, and execution. The researchers concluded that AI would be useful in wargaming in narrow and more specialized applications like tactical gameplay and providing insights into wargaming results. But current AI exposes wargaming to risk such as sensitive data breaches or misinformation stemming from faulty gaming logic. These risks will need to be managed. And AI is currently inadequate for use in more complex areas of wargaming such as course of action analysis and adjudication. In short, the benefits of AI in wargaming are limited and the risks great. Current AI is better utilized in simulations of systems rather than in attempts to replace or evaluate human decision.

?

Simulations: The military uses simulations in various ways to improve force readiness. Combat training can be made more realistic while enhancing safety and lowering costs. AI can be used to generate and adapt scenarios that challenge individuals and units with constantly evolving problem sets. Battle drills can be automated allowing soldiers to hone critical skills. Protocols, maneuvers, and weapons handling can be practiced with safety. Feedback can be automated and rapidly provided to commanders and trainees.

Another area in which simulations can be useful is in mission rehearsal. The details of planning and coordination can be considered in a training environment that closely resembles reality in a cognitive sense. Mission execution can be iteratively repeated with AI helping capture lessons learned so they can be turned into improvements in execution. Aspects of mission execution such as insertion, battle space and fires coordination, and air asset control can be simulated and measured.

Simulations can also add in logistics and administration, all key to mission success. As the saying “amateurs study tactics but professionals study logistics” attributed the Gen. Robert H. Barrow, Commandant of the United States Marine Corps in 1980 suggests, it is arguably more important to integrate AI into the development and training of logistical and administrative systems than into tactical systems. The quantitative aspects of logistics and administration can facilitate AI exploitation but, as always, in a human-machine teaming environment. ??

?

Experimentation: Another valuable tool in training and development is experimentation. While similar to simulations, experimentation differs in that simulations generally are computer-generated models that mimic reality. Experimentation refers to tests or studies conducted in the real world to prove or confirm hypotheses. Simulations emulate reality while experimentation proves the validity of concepts. Simulations can often be performed with minimal risk while experimentation often requires acceptable risk based on cost return analysis and ethics.

?

Experimentation can be used to test new technologies, improve strategic planning, and support adaptation to the changing operational environment. It can influence doctrine, force structure, and procurement. It is a critically important tool that aids innovation, decision-making, and operational effectiveness.

?

Conclusions

?

????????????? Artificial intelligence is transforming the training and development environment. As this technology advances and is able to interact with humans in increasingly complex and realistic ways, AI will begin to play a greater role in all aspects of the human experience. The military is leveraging the methodologies of AI in many ways to increase operational effectiveness. As it does this, the military is becoming aware of the limitations of AI. Experience shows that AI is still far from surpassing human intelligence. As activities become more subjective and relatively less quantitative, the advantages AI has over human intelligence, such as speed of information processing and endurance, become less important than the unparalleled ability of humans to react to their environment in new ways and to appreciate unforeseen linkages in information. This suggests that the way ahead is in human-machine teaming that leverages the unique strengths of AI and human intelligence in a complementary manner. AI can improve training, wargaming, simulations, and experimentation. But it will require resourcing, adaptation, and a willingness by military culture to embrace ambiguity and innovation. There will be a tendency for modern bureaucratic militaries to delay or reject the leveraging of AI. This is dangerous as the military that best integrates AI into its doctrine and force structure will likely dominate in future conflict.

[i] West, D. and Allen, J. (2018). How Artificial Intelligence is Transforming the World. Brookings Institute. Washington, DC. https://www.brookings.edu/articles/how-artificial-intelligence-is-transforming-the-world/

[ii] Coursera Staff (2024). What is Artificial Intelligence? Definitions, Uses, and Types. Mountain View, CA. https://www.coursera.org/articles/what-is-artificial-intelligence

[iii] Hintze, A. (2016). Understanding the Four Types of Artificial Intelligence. Government Technology. e.Republic. https://www.govtech.com/computing/understanding-the-four-types-of-artificial-intelligence.html

[iv] Sentient Digital, Inc. (2024). The most useful Applications of AI by the Military in 2024 and beyond. New Orleans, LA. https://sdi.ai/blog/the-most-useful-military-applications-of-ai/

[v] Nurkin, T, and Siegel, J. (2023). How Modern Militaries are Leveraging AI. Atlantic Council. Washington, DC. https://www.atlanticcouncil.org/about/visiting-atlantic-council/

[vi] Department of Defense (2022). Summary of the Joint All-Domain Command and Control (JADC2) Strategy. Washinton DC. to seize, maintain, and protect our information and decision advantage over our adversaries.

[vii] Miller, A. (2021). AI Algorithms deployed in Kill Chain Target Recognition. Air and Space Force Magazine. Arlington, VA. https://www.airandspaceforces.com/ai-algorithms-deployed-in-kill-chain-target-recognition/

[viii] OpenAI (2021). GPT-3 Powers the Next Generation of Apps. Fable Studio, San Francisco, CA. https://openai.com/about/

[ix] Google. PaLM 2. https://ai.google/discover/palm2/

[x] Allen, J. and Husain, A. (2017). On Hyperwar. Proceedings. Annapolis, MD. https://www.usni.org/magazines/proceedings/2017/july/hyperwar

[xi] Department of Defense (2023). Data, Analytics, and Artificial Intelligence Adoption Strategy. https://media.defense.gov/2023/Nov/02/2003333300/-1/-1/1/DOD_DATA_ANALYTICS_AI_ADOPTION_STRATEGY.PDF

[xii] Mission critical learning at the speed of need | Cerego

[xiii] Jordan, S. (2018). Army invests in Predictive Maintenance for Bradleys. National Defense. Arlington, VA. https://www.nationaldefensemagazine.org/articles/2018/9/26/army-investing-in-predictive-maintenance-for-bradleys

[xiv] Kushner, N. (2024). As Ukraine focuses on drone warfare, its military creates new Unmanned Systems Forces branch. ABC News. New York, NY. https://abcnews.go.com/International/ukraine-focuses-drone-warfare-military-creates-new-unmanned/story?id=111046806

[xv] Alotaibi, H., Salamah, D (2023). The impact of translation apps on translation students’ performance.?Educ Inf Technol?28, 10709–10729. https://doi.org/10.1007/s10639-023-11578-y

[xvi] Wright, L. (2023). Artificial intelligence and its role in Training and Development. London, England. Artificial Intelligence and its Role in Employee Training and Development - Businesstechweekly.com

[xvii] Menthe, Lance, Li Ang Zhang, Edward Geist, Joshua Steier, Aaron B. Frank, Erik Van Hegewald, Gary J. Briggs, Keller Scholl, Yusuf Ashpari, and Anthony Jacques, Understanding the Limits of Artificial Intelligence for Warfighters: Volume 1, Summary. Santa Monica, CA: RAND Corporation, 2024. https://www.rand.org/pubs/research_reports/RRA1722-1.html. Also available in print form.

[xviii] Knack, A. and Powell, R. (2023). Artificial Intelligence in Wargaming. Centre for Emerging Technology and Security. London, UK. https://cetas.turing.ac.uk/sites/default/files/2023-06/cetas_research_report_-_ai_in_wargaming.pdf